MLflow Skills
Closes the loop for agent work: instrument, trace, evaluate, change, then verify the change helped.
Install
OfficialShips scriptsClaude Code: /plugin marketplace add anthropics/claude-plugins-official → /plugin install mlflow@claude-plugins-official
Bir LLM uygulamasına izleme eklenirken, iz üzerinden arıza aranırken ya da değerlendirme puanlayıcısı kurulurken
- Author
- MLflow
- License
- Apache-2.0
The pieces are ordinary MLflow — tracing for Python and TypeScript, trace retrieval, metric queries, failure analysis — but the arrangement is the point, because without a scorer an agent improvement is just an opinion. The scorer skill refuses to start from the catalogue: it works out what the application does, extracts a small set of atomic quality criteria from that, and only then picks the cheapest scorer that can judge each one, treating the user as the authority on what counts as wrong. Aiming for a runnable and inspectable prototype rather than a perfect first pass is what keeps the loop turning.
Similar skills
Skill Seekers
SkillTurns documentation sites, repos and PDFs into agent skills.
PluginEval
SkillScores a Claude Code skill on ten dimensions across static checks, an LLM judge and Monte Carlo runs.
DeepEval Skills
SkillEvaluation, tracing and OpenTelemetry export for LLM apps, split into three non-overlapping skills.
Comments(0)
Sign in to comment